Table 6.

Performance of multiclass classification strategies – one shot classifier (OSC), one vs. all (OVA), cascaded classifier in CaMCCo – upon fusion of modalities chosen from training set for each classification task.

		ACC	BACC	AUC	SEN	SPEC	PPV
CN	${\mathscr{H}}_{OSC}$	0.69	0.63	0.96	0.34	0.92	0.75
	${\mathscr{H}}_{OVA}$	0.89	0.77	0.97	0.59	0.96	0.77
	${\mathscr{H}}_{CAMCCO}$	0.89	0.77	0.97	0.59	0.96	0.77
MCI	${\mathscr{H}}_{OSC}$	0.69	0.69	0.77	0.68	0.70	0.67
	${\mathscr{H}}_{OVA}$	0.68	0.68	0.77	0.78	0.59	0.63
	${\mathscr{H}}_{CAMCCO}$	0.80	0.78	0.89	0.88	0.69	0.80
AD	${\mathscr{H}}_{OSC}$	0.69	0.67	0.84	0.53	0.82	0.69
	${\mathscr{H}}_{OVA}$	0.85	0.82	0.90	0.72	0.91	0.81
	${\mathscr{H}}_{CAMCCO}$	0.80	0.78	0.89	0.69	0.88	0.80

The highest accuracy (ACC), balanced accuracy (BACC), area under the ROC curve (AUC), sensitivity (SEN), specificity (SPE) and positive predictive value (PPV) achieved for each class are shown in bold. These results indicate that although the performance of both ${\mathscr{H}}_{OVA}$ and ${\mathscr{H}}_{CAMCCO}$ are comparable for CN and AD classification, ${\mathscr{H}}_{CAMCCO}$ outperforms all other methods for MCI classification.